Altered sleep during spontaneous cannabinoid withdrawal in male mice

Abstract

Cessation of cannabinoid use in humans often leads to a withdrawal state that includes sleep disruption. Despite important health implications, little is known about how cannabinoid abstention affects sleep architecture, in part because spontaneous cannabinoid withdrawal is difficult to model in animals. In concurrent work we report that repeated administration of the high-efficacy cannabinoid 1 (CB1) receptor agonist AM2389 to mice for five days led to heightened locomotor activity and paw tremor following treatment discontinuation, potentially indicative of spontaneous cannabinoid withdrawal. Here, we performed parallel studies to examine effects on sleep. Using implantable electroencephalography (EEG) and electromyography (EMG) telemetry we examined sleep and neurophysiological measures before, during, and after five days of twice-daily AM2389 injections. We report that AM2389 produces decreases in locomotor activity that wane with repeated treatment, whereas discontinuation produces rebound increases in activity that persist for several days. Likewise, AM2389 initially produces profound increases in slow-wave sleep (SWS) and decreases in rapid eye movement (REM) sleep, as well as consolidation of sleep. By the third AM2389 treatment, this pattern transitions to decreases in SWS and total time sleeping. This pattern persists following AM2389 discontinuation and is accompanied by emergence of sleep fragmentation. Double-labeling immunohistochemistry for hypocretin/orexin (a sleep-regulating peptide) and c-Fos (a neuronal activity marker) in lateral hypothalamus revealed decreases in c-Fos/orexin+ cells following acute AM2389 and increases following discontinuation, aligning with the sleep changes. These findings indicate that AM2389 profoundly alters sleep in mice and suggest that sleep disruption following treatment cessation reflects spontaneous cannabinoid withdrawal.

INTRODUCTION

Cannabis products are widely consumed, in part with the expectation that they will improve sleep (Winiger et al., 2021). The effects of cannabinoids on sleep are well-documented in research literature (Kesner and Lovinger, 2020). For example, acute administration of the CB1 partial agonist Δ⁹-tetrahydrocannabinol (THC; the major psychoactive component of cannabis) in early clinical sleep studies was shown to decrease the latencies for both sleep onset and waking after sleep onset (Cousens and DiMascio, 1973; Feinberg et al., 1975). In a more recent study, 71% of participants cited improved sleep as a reason for the use of medicinal cannabis (Bonn-Miller et al., 2014). Realization of the strong hypnogenic effects of cannabis and cannabinoid drugs has led to a growing interest in their potential utility in the treatment of various sleep disorders including insomnia or nightmares associated with post-traumatic stress disorder (Kaul et al., 2021). Surprisingly, the effects of repeated or chronic cannabinoid use on sleep measures currently are not well understood. Small studies in which sleep is evaluated as a secondary measure have shown a small benefit of cannabinoids on self-reported sleep (Kuhathasan et al., 2019). However, data from large well-controlled studies using objective measures to document cannabinoid effects on sleep are not yet available (Suraev et al., 2020). Evidence suggests that tolerance to the sleep-enhancing effects of cannabinoids can occur quickly, which may complicate their potential benefit in the treatment of sleep disorders (Angarita et al., 2016).

Cessation from long-term cannabinoid use often precipitates a specific set of symptoms known as cannabis withdrawal syndrome, and sleep disturbances are one of six diagnostic criteria for this condition (Bonnet and Preuss, 2017). A meta-analysis found that 47% of regular cannabis users exhibited cannabis withdrawal syndrome following discontinuation of use (Bahji et al., 2020). A systematic review of cannabis withdrawal and sleep notes that, while sleep disruption following withdrawal appears to be common across multiple studies, small sample sizes and the presence of confounding variables limit determination of the specific aspects of sleep that are affected (Gates et al., 2016). The authors of that review concluded that well-controlled studies in both clinical and laboratory settings are needed to more thoroughly characterize how chronic cannabinoid use alters sleep architecture.

Prior studies have shown that cannabinoid drugs can reliably alter sleep architecture in both humans and laboratory animals. Acute exposure to THC typically produces a pattern of increased sleep consolidation, increased slow-wave sleep (SWS), and decreased rapid-eye movement sleep (REM) (Kaul et al., 2021; Kesner and Lovinger, 2020). Results from studies of chronic cannabinoid exposure are variable but there is growing evidence for decreased total sleep time, decreased SWS, and a return to baseline levels of REM (Kaul et al., 2021; Kesner and Lovinger, 2020). Studies investigating how sleep is affected during cannabinoid withdrawal are even more limited. Sleep abnormalities associated with disrupted cannabis use include subjective measures such as increased sleep disturbances and vivid dreams, coinciding with changes of objective measures such as reduced sleep efficiency, total sleep time, and SWS, and increased REM (Angarita et al., 2016). While animal models are often used in objective and controlled studies of drug dependence and withdrawal, a major barrier to understanding how cannabinoid withdrawal alters sleep is the lack of distinct and reproducible signs of spontaneous cannabinoid withdrawal in frequently studied species such as rats and mice.

In a corresponding study reported in this issue (Paronis et al., 2021), we established dosing regimens in mice that produce physiological signs of cannabinoid withdrawal following the cessation of repeated treatment with the cannabinoid AM2389 (Jarbe et al., 2012; Paronis et al., 2013). Unlike THC, AM2389 is considered to be a full agonist at the CB1 receptor and was used in the present studies with the expectation that exposure to an agonist with efficacy higher than that of THC would result in correspondingly greater dependence and more pronounced signs of drug withdrawal. We found that daily administration of AM2389 to mice for five days produced observable signs of spontaneous withdrawal characterized by increased paw tremors and increased locomotor activity at 12-48 hours after cessation of treatment. Comparing these outcomes with data from THC-treated mice demonstrated that the effects of AM2389 were quantitatively greater on all measures. In the study reported here, we investigated whether sleep and other sleep-related metrics are altered by the same five-day treatment regimen of AM2389. In male mice implanted with wireless electroencephalographic (EEG) and electromyographic (EMG) telemetry devices we continuously monitored sleep before, during, and following AM2389 treatment, and compared multiple metrics within individual subjects. We report that in addition to the previously identified alteration in locomotor activity, treatment with the high efficacy CB1 agonist produces profound effects on sleep and EEG spectral power over the course of treatment with AM2389 and following its cessation.

METHODS

Animals

Male C57BL/6J mice, 8-10 weeks old, 21-26g bodyweight, were obtained from Jackson Laboratory (Bar Harbor, ME, USA). Following subcutaneous implantation with transmitters and seven days of recovery, mice were transferred to an isolated housing room that was temperature (21 ± 2°C and humidity (50 ± 20%) controlled under a 12:12 hr light dark schedule. Food and water were available ad libitum throughout the duration of the experiment. Mice were habituated to the room for seven days prior to starting the experiment. All experimental procedures were approved by the Institutional Animal Care and Use Committee at McLean Hospital under protocol 2017N000091.

Experimental procedure

Adult male mice were implanted with telemetry devices to enable continuous collection of EEG, EMG, locomotor activity and subcutaneous temperature measures over the course of the experiment, as described previously (Missig et al., 2018; Wells et al., 2017). After 7 days of recovery from transmitter implantation procedures, two 24-hour periods of baseline were recorded. Following baseline, AM2389 (0.03 mg/kg, subcutaneous injection at 10ml/kg) or saline (10ml/kg) was administered at lights-off (19:00) and again at lights-on (07:00). These twice-daily injections were repeated for five consecutive days (D1-D5) (Figure 1). As the first injection began at 19:00 this was considered the start of each 24-hr experiment day. On the last day one injection was administered at 19:00, such that this period (A0) represents the first 24 hours following the last injection. Mice were monitored for 7 more days during abstinence from the drug (A1-A7). This experimental regimen was identical to that described in the concurrent report (Paronis et al., 2021). Activity and temperature were analyzed throughout the entire experiment, whereas sleep and EEG spectral power were analyzed for a subset of days. An additional cohort of mice were perfused on either D1 or A1 for immunohistochemistry (Figure 1).

Fig. 1: Schematic of experiment design.

Following two baseline days, AM2389 or saline was administered for five days (D1-D5), both during lights on (0700) and again at lights off (1900). Data recordings continued for eight days following cessation of treatment (A0-A7), with A0 representing the first 24 hours following last treatment. Sleep and EEG metrics were analyzed for a subset of days, whereas locomotor activity and temperature were analyzed throughout the experiment. Separate groups of mice were sacrificed on D1 (dark phase) or A1 (light phase) for immunohistochemistry.

Transmitter implantation

Electroencephalogram (EEG) and electromyogram (EMG) leads from a radiotelemetry transmitter (F20-EET, Data Science International [DSI], St Paul, MN, USA) were implanted under ketamine/xylazine (100/10 mg/kg, IP) anesthesia as described previously (Missig et al., 2018; Wells et al., 2017). Mice were stereotaxically immobilized and an incision was made from the medial skull to the posterior portion of the neck. Transmitters were implanted subcutaneously between the forelimb and the hindlimb. Two insulated biopotential wires (EMG electrodes) were inserted into the trapezius muscle and held in place using non-dissolvable sutures. Next, two stainless steel screws were inserted into the skull, serving as EEG electrodes (relative to bregma: AP +1.0 mm, ML +1.0 mm, and AP −3.0 mm, ML −3.0 mm). The screws and EEG leads were held in place with dental cement over the top of the skull. Mice received a post-surgery analgesic (ketoprofen, 5 mg/kg, IP) once, 24 hours following surgery. They were singly-housed in a standard Plexiglas home cage for at least seven days prior to the start of the experiment.

Data acquisition

Individual mouse cages were situated on top of RPC-1 PhysioTel receivers (DSI), which detect signals from the transmitters. The receivers were connected to a data exchange matrix (DSI) that continuously uploaded recording data to a computer at a sampling rate of 250 Hz for EEG and EMG data, and 1 Hz for temperature and locomotor activity data. Locomotor activity and temperature data were collected using Ponemah Software (DSI) whereas EEG and EMG data were analyzed using Neuroscore (DSI).

Analysis of EEG/EMG data

Data were first mean-corrected and epoched into 10 second bins. A Fast-Fourier Transformation (FFT) was performed on each bin to translate the data into frequency space. EEG spectral power frequency bands were defined as Delta (0.5-4 Hz), Theta (4-8 Hz), Alpha (8-12 Hz), Beta (16-24 Hz), and Gamma (30-50 Hz). Mouse vigilance states were then manually scored by a blinded human experimenter and classified as three distinct phases: Wake (W), Slow-wave sleep (SWS), and REM sleep (REM). The vigilance state of each 10s epoch was classified using the following metrics; W was defined by body and limb movements as evident by elevated EMG amplitude and/or presence of locomotor activity, and the presence of low-voltage EEG amplitudes; SWS was defined by a lack of body movements, low amplitude EMG activity, and an emergence of high amplitude delta waves (0.5-4 Hz) in the EEG; and REM was defined by a similar low amplitude EMG activity coupled with the presence of theta wave (4-8 HZ) in the EEG. EEG artifacts were characterized by extremely high amplitude non-physiological signals and were excluded from analysis of EEG spectral power; however, they were treated as the same sleep stage as the previous epoch for the purposes of sleep bout analysis. Sleep bouts were defined as any sleep stage that persisted for two or more consecutive epochs (20s or longer) and were considered over when interrupted by an epoch of a different sleep stage. Eight representative 24-hour periods from 7:00 pm - 7:00 pm were scored for each mouse; these periods were selected based on locomotor activity data as well as results from a companion study (Paronis et al., 2021) and included baseline, three drug days (D1, D3, D5), a transition day (A0), and three post-drug days (A1, A2, A5). These days were chosen to provide information on both the acute and repeated effects of drug administration and examine both the immediate and later effects following cessation. Sleep bout distribution was analyzed by examining three bins representing approximately 20-40% of total time at baseline following a log₂ progression to enable examination of sleep bout distribution over repeated days.

Analysis of locomotor activity and body temperature

Measures of locomotor activity and temperature were made using the same F20-EET transmitters. In this system, locomotor activity was derived from signal strength variation from the telemetry implant as the mice moved across the receiver in their home cage and is related to both distance and speed of movement. Temperature was collected from a sensor within the transmitter. Locomotor activity and body temperature were recorded continuously throughout the experiment and the mean temperature and sum of activity were analyzed in 1-hr bins.

Immunohistochemistry

Immunohistochemistry was performed in a similar manner as described (Missig et al., 2020) (Li et al., 2018). Mice were deeply anesthetized and transcardially-perfused with phosphate buffered saline, followed by 4% paraformaldehyde. Brains were dissected, post-fixed in 4% paraformaldehyde overnight at 4°C, equilibrated to 30% sucrose, and then cryosectioned (30μm). Sections containing the lateral hypothalamus (LH) were permeabilized with 0.3% Triton X-100, blocked with 1% bovine serum albumin, and then incubated with 1:500 rabbit anti-c-Fos antibody (sc-52 Santa Cruz Biosciences) and 1:1000 goat-anti-Orexin antibody (sc-8070, Santa Cruz Biosciences) overnight at 4°C. After 3 rinses, sections were incubated in 1:400 AlexaFluor 488 and 555 antibodies for 2 hours, rinsed again, and then mounted on slides. Images of the LH were obtained with a 20x objective using a confocal laser scanning microscope (TCS SP8, Leica, Wetzler, Germany). All images were acquired under identical settings and parameters, and then exported to NIH ImageJ for analysis. Colocalization of hypocretin/orexin and c-Fos was performed by a blinded scorer using ImageJ.

Statistical Analysis

Data were analyzed using mixed-effects models with repeated measures comparing (treatment X day [repeated]). Contrasts were made comparing saline to AM2389 for each day and corrected using Sidak’s correction for multiple comparisons. A False Discovery Rate (FDR) using a two-stage set-up method was performed for locomotor activity and body temperature due to numerous comparisons being made (n=360; (Yoav Benjamini, 2006). Immunohistochemistry endpoints were compared using t-tests. Normality of data was assessed via F tests or analysis with QQ plots. Two saline-treated mice had EEG leads fail on withdrawal day 0 and both sleep and spectral analysis for this day and subsequent days were excluded from analysis. Statistical analyses were performed with GraphPad Prism 8.1 and significance for all analyses was set at p<0.05.

RESULTS

AM2389 treatment decreases locomotor activity and body temperature during administration, but increases locomotor activity following discontinuation of treatment.

AM2389 induced an almost complete suppression of locomotor activity initially, especially during the dark phase of D1 (19:00-23:00 and 04:00-07:00), which partially continued in the dark phases of D2 (19:00-22:00 and 00:00), and on D3 (23:00) (Figure 2A). Following the initial suppression of locomotor activity, increases in locomotor activity emerged during the transition from dark to light phase and steadily increased over drug exposure days (D3-A0). Following discontinuation of AM2389 treatment there was a gradual elevation of locomotor activity for several days, especially during the dark phase, with significant increases at numerous time points throughout A1-A7. Locomotor activity in AM2389-treated mice exhibited the same polyphasic diurnal pattern as saline-treated mice, but with higher magnitude of peak activity levels. A substantial decrease in subcutaneous body temperature coincided with the initial decrease in locomotor activity on D1 (Figure 2B). The first injection on D1 led to a significant decrease in temperature to a nadir of 4-5 degrees less than saline treated mice (20:00-16:00). On D2 temperatures initially decreased during the dark phase (20:00-04:00), but then began to gradually normalize during the light phase.

Fig. 2. Repeated AM2389 treatment alters locomotor activity and temperature.

Locomotor and temperature were recorded throughout the experiment using a subcutaneous transmitter. A) Locomotor activity is suppressed following injection of AM2389 and exhibits a gradual enhancement several days following cessation. B) Subcutaneous body temperature is markedly suppressed for the first 48 hours during repeated AM2389 administration before normalizing. Solid and dotted lines represent the means ± SEM (shaded region) of n = 7 per group, yellow markers indicate p < 0.05 FDR corrected (vs. Saline).

AM2389 alters sleep patterns during and following cessation of treatment

To determine how repeated AM2389 treatment affects sleep, each 10-s epoch was classified as either Wake (W), Slow-wave Sleep (SWS), or Rapid-Eye Movement Sleep (REM). AM2389 induced an initial increase of SWS on D1 (Figure 3B), with a concordant decrease in W (Figure 3A). The increase in SWS and decrease in W occurred only during the initial dark phase following the first drug administration on D1 (Figure 3D,E), and there were no significant changes during the following light phase (Figure 3G,H). By D3 an inverse pattern emerged with AM2389 resulting in a decrease in total SWS (Figure 3B) and an increase in W (Figure 3A); this pattern continued throughout the drug exposure regimen and for several days after discontinuation of AM2389 treatment. The decrease in SWS and increase in W occurred primarily during the light phase on D3, D5, A0 (Figure 3G,H); although following discontinuation of AM2389 treatment, SWS also was decreased on A1 and A2 during the dark phase (Figure 3D,E). Overall, these data show that AM2389 initially promoted SWS; however continuous, repeated exposure to the CB1 agonist led to a suppression of SWS and increase in W that persisted into the first several days of abstinence. In contrast, REM was almost completely eliminated following the initial administration of AM2389 on D1, with only a slight recovery on D3 and D5 (Figure 3C,F,I). However, REM quickly returned to control levels following cessation.

Fig. 3: Repeated AM2389 treatment alters sleep.

AM2389 produced significant alterations in percent time animals spent in Wake (W), Slow-wave Sleep (SWS), and REM sleep during repeated administration and following cessation. Vigilance state over 24 hours for A) Wake, B) SWS, and C) REM, showing initial increase in SWS on D1 transforming into a decrease by D3 that continued through A2. Examining data only during the 12hr (D-F) dark phase or (G-I) light phase revealed that initial increase in SWS on D1 is primarily during the dark phase, but the subsequent decrease in SWS from D3-A0 is during the light phase. During A1 and A2 both light and dark phases exhibit a decrease in SWS. * p < 0.05, ** p < 0.01 (vs. Saline), Mean ± SEM, n = 7 (AM2389) and n=5-7 (saline).

Cessation from AM2389 treatment is characterized by fragmented sleep

Analysis of sleep bouts can be used to determine changes in the structure of sleep patterns. Here, the length of bouts (uninterrupted, consecutive epochs) of each vigilance state was calculated and the total proportion of time spent in different bout durations was separated into three bins, each representing ~20-40% of total time at baseline. Administration of AM2389 led to an apparent consolidation of SWS on D1, characterized by a significant increase in longer SWS bouts (≥640s)(Figure 4B) and a decrease in longer W bouts (≥2560s)(Figure 4A). However, this consolidation was transient, with SWS returning toward baseline values on D3 of the 5-day treatment regimen.

Fig. 4: Repeated AM2389 affects sleep architecture.

The length of bouts (uninterrupted, consecutive epochs) of each vigilance state were calculated and the total proportion of time spent in different bout durations was separated into three bins each representing ~20-40% of total time at baseline and divided into A) W, B) SWS, and C) REM sleep. No differences reached significance in saline treated animals. On D1, AM2389 lead to a consolidation of SWS with increases in longer bouts, and increase in shorter bouts of W. REM sleep on D1 is omitted as almost no REM sleep was observed during this period (Fig. 3). Following cessation of AM2389 there is an increase in shorter bouts of SWS that peaks on A1. * p < 0.05 ** p < 0.01 (vs. Saline), Mean ± SEM, n = 7 (AM2389) and n=5-7 (saline).

Data obtained after the repeated regimen ended suggests that SWS was fragmented for multiple days following the cessation of treatment with AM2389. Thus, a prominent increase in shorter bouts of SWS (<320s) was evident on A0, peaked on A1, and returned toward control values on A5 following the cessation of AM2389 treatment. While there were numerous trends, there were no statistically significant alterations in REM sleep bouts throughout the study, particularly since D1 was excluded due to negligible time in REM (Figure 4C).

AM2389 changes EEG spectral power during and following cessation from treatment

Alterations in EEG spectral power may reflect global changes in neuronal activity patterns. Changes in EEG spectral power from baseline of each mouse were calculated for each vigilance state. For statistical analysis, absolute EEG powerbands were analyzed by comparisons to baseline (Figure 5B, D, F). Overall change using a Fast-Fourier Transformation of spectral power (non-binned) is also shown for D1 and A1 (Figure 5A, C, E). AM2389 produced an initial decrease on D1 in EEG power preferentially in higher frequency powerbands in W (Gamma) and SWS (Alpha, Beta, Gamma); note that REM on D1 was excluded due to negligible time in this vigilance state. One exception is a prominent increase in 3-6 Hz EEG spectral power during W (Theta) (Figure 5A,B). These changes diminish over consecutive days of AM2389 treatment, with no significant differences detectable by D3 or D5. However, following discontinuation of AM2389 treatment, an inverse pattern emerged, with an overall increase in spectral power in higher frequency powerbands on A1 for W (Theta, Alpha, Beta, Gamma) (Figure 5B), SWS (Gamma) (Figure 5D), and REM (Alpha, Beta, Gamma) (Figure 5F). A similar pattern of increased EEG spectral power was observed on A2 and diminished slightly by A5, although significant increases remained in several bands.

Fig. 5: Repeated AM2389 alters EEG spectral power.

EEG spectral power was analyzed per epoch and analyzed per vigilance state and compared to baseline for each animal. A,C,E) change from baseline using Fast Fourier Transformation is shown above for D1 and A1, and B,D,F) power binned into powerbands of Delta (0.5-4 Hz), Theta (4-8 Hz), Alpha (8-12 Hz), Beta (16-24 Hz), Gamma (30-50 Hz) for statistical analysis is shown below. REM sleep on D1 is omitted as almost no REM sleep was observed during this period (Fig. 3). Overall, AM2389 resulted in an initial in decrease in EEG power preferentially in higher frequency powerbands during D1 and an increase following cessation. * p < 0.05 ** p < 0.01 (vs. Saline), Mean ± SEM, n = 7 (AM2389) and n=5-7 (saline).

Activation of Orexin neurons during and following cessation of AM2389 treatment

As AM2389 profoundly altered sleep, we next investigated the possibility that this could be mediated through direct effects on sleep/wake circuitry. Orexin-expressing neurons within the lateral hypothalamus are well-established as critical neurons in promoting wakefulness (Estabrooke et al., 2001). To determine whether AM2389 affects activity of orexin neurons, a separate cohort of mice was rapidly euthanized 2 hours following a single injection of 0.03 mg/kg AM2389 (D1) during the active period (at 21:00), a time when orexin neuronal activity is high under normal conditions. Another cohort was euthanized following treatment cessation at 09:00 of A1, when sleep drive is high and orexin neuronal activity is low under normal conditions (Estabrooke et al., 2001). The percentage of orexin neurons that co-labeled with the immediate early gene c-Fos was quantified (Figure 6A). Acute administration of AM2389 led to a significant decrease in c-Fos+ orexin neurons (Figure 6B), consistent with the decrease in wakefulness during this period and suggesting that CB1 signaling may inhibit activity of these neurons. Conversely, following discontinuation of AM2389 treatment (on A1), there were significant increases in c-Fos+ orexin neurons (Figure 6C) consistent with increases in wakefulness observed during this period and suggesting that cessation of cannabinoid treatment leads to aberrant activation of the orexin wake-promoting neurons.

Fig. 6: Repeated AM2389 affects activity of sleep/wake circuitry.

Double-labeling immunohistochemistry for c-Fos and Orexin in the lateral hypothalamus was examined as a well-characterized neuronal population that promotes wakefulness. A) example images of orexin, c-Fos, and combined at 200X magnification with inset on bottom right showing Orexin and c-Fos double-labeling. B) AM2389 decreased the percent of orexin neurons expressing c-Fos on D1 during the dark phase, but C) increased the percent of orexin neurons expressing c-Fos on A1. * p < 0.05 ** p < 0.01 (vs. Saline), Mean ± SEM, n = 4-6 per group.

DISCUSSION

The present results show that the CB1 full agonist AM2389 can markedly alter locomotor activity, body temperature, sleep, and EEG spectral power in mice, and that its effects on these endpoints change over the course of multiple days of treatment and following treatment cessation. Initially (first 24 hr), treatment with AM2389 results in suppression of locomotor activity and a dramatic decrease in body temperature. The changes in behavior and temperature coincide with an increase in, and consolidation of, SWS; at the same time, there is an almost complete loss of REM sleep. The initial injection of AM2389 decreases EEG spectral power, especially higher frequencies, with the sole exception of a prominent increase in theta power during wakefulness. Coinciding with these changes is a decrease in c-Fos+ orexin neurons in the lateral hypothalamus, indicative of lower activity of wake-promoting neurocircuitry. After these initial effects, during days 3-5 of treatment with AM2389, locomotor activity, body temperature, and EEG spectral power began to return to near-baseline levels. However, sleep patterns remain disrupted during this period, with a notable decrease in SWS and an increase in wakefulness. A distinct pattern of neurophysiological alterations emerges during abstinence from AM2389 treatment. Locomotor activity is elevated for up to seven days following cessation, with no major changes in body temperature. SWS remains decreased and wakefulness increased, especially during the dark phase, however, the fragmentation of SWS is perhaps the most notable change following the cessation of AM2389 treatment. This fragmentation peaks during the first 24-72 hours following discontinuation of AM2389 treatment, with ≤50% of all bouts lasting longer than 320 s. EEG spectral power, especially in the higher frequencies, was increased during W and SWS for at least five days following treatment cessation. Paralleling these changes was an increase in the number of c-Fos+ orexin neurons during the light phase during abstinence, suggesting an enhancement in the activity of wake-promoting orexin neurocircuitry. Overall, these results demonstrate that repeated cannabinoid exposure can result in tolerance to cannabinoid-related changes in sleep-related metrics as well as signs of spontaneous withdrawal during abstinence that include decreases in sleep and sleep fragmentation.

In this study, we replicated the findings of activity and temperature changes that occur following a regimen of repeated AM2389, which we describe in a concurrent report (Paronis et al., 2021). In both studies, locomotor activity was suppressed for the first 48 hours of AM2389 treatment and returned to near-baseline levels on D5. Hourly analysis of the data in the present study further clarifies that the recovery of baseline activity levels results from an altered diurnal activity pattern, with peaks in locomotor activity occurring at the end of the dark phase. This pattern may reflect a compensatory increase in locomotor activity when drug levels are lowest and suggests that AM2389 continues to have sustained suppressive effects on locomotor activity. After discontinuation of the AM2389 treatment, the pattern of locomotor activity briefly returned to normal diurnal activity patterns before another new pattern of elevated total activity levels emerges—characterized by elevated total activity levels especially during A3-A7—likely indicative of spontaneous cannabinoid withdrawal. Also consistent with the concurrent report (Paronis et al., 2021), body temperature decreased during the first two days of AM 2389 treatment then returned toward vehicle levels upon discontinuation. Unlike other endpoints, temperature was not profoundly altered during the withdrawal period, demonstrating that not all cannabinergic endpoints undergo rebound effects following discontinuation of treatment. Overall, these findings suggest that repeated AM2389 administration reliably produces a pattern of locomotor activity changes that can be associated with spontaneous cannabinoid withdrawal.

The manner in which AM2389 affected sleep is broadly consistent with what is known about acute cannabinoid exposure and, further, provides evidence that sleep fragmentation may occur as part of withdrawal. Studies of THC or other cannabinoids in animal models and in humans generally report an enhancement of SWS, suppression of REM, and/or decreases in sleep onset latency (Kaul et al., 2021; Kesner and Lovinger, 2020). In contrast, studies of prolonged exposure to cannabinoids in regular cannabis users and, to a very limited extent, laboratory animals suggest that such exposure can lead to impairment in sleep, with selective and persistent decreases in SWS and initial suppression of REM (Kaul et al., 2021; Kesner and Lovinger, 2020). This pattern is recapitulated in the present study of AM2389 exposure, with an enhancement of SWS and initial suppression of REM, that transitioned into decreased SWS by the third injection day, accompanied by reductions in REM suppression. Although sleep disturbances are one of the hallmark symptoms of cannabinoid withdrawal in humans (Kesner and Lovinger, 2021), there are few reports in which changes in sleep architecture following cannabinoid cessation are described. Published reports of EEG sleep measures following cessation of cannabis use consistently demonstrate decreases in SWS relative to nondrug-experienced control subjects, whereas changes in REM are more variable (Bolla et al., 2008; Bolla et al., 2010; Vandrey et al., 2011). Here, we found that following discontinuation of AM2389 treatment the levels of SWS remained low for at least five days, even as REM returns to baseline levels. Most strikingly, a fragmentation of SWS—indicated by an increase in shorter bouts of SWS—emerged over the first 3 days following treatment discontinuation. The fragmentation of SWS peaked during the first 24-48 hours following treatment discontinuation and approximates baseline and saline values within five days post-treatment. Consolidation of SWS following initial exposure and the fragmentation of SWS following cessation of AM2389 injections in mice is consistent with similar to patterns previously reported for human and non-human subjects (Kaul et al., 2021).

The alterations in sleep during and following repeated AM2389 correlate with altered activity in lateral hypothalamic hypocretin/orexin neurons, which play a well-characterized role in promoting wakefulness (Carter et al., 2009; de Lecea, 2021; Estabrooke et al., 2001). As anticipated, after initial exposure to AM2389 during the dark phase—when orexin neuron activity is expected to be higher—c-Fos labeling in orexin neurons was decreased compared to controls corresponding to the increase of SWS. Similarly, following discontinuation of AM2389 treatment during the light phase—when orexin neuron activity is expected to be lower—c-Fos labeling in orexin neurons was increased, aligning with the decrease in SWS and increased sleep fragmentation observed during this period. These findings suggest that cannabinoid withdrawal causes changes in sleep through affecting activity of neuronal sleep circuitry. It is important to note, however, that these studies are correlational and direct causal studies using targeted neuronal manipulations are needed to fully determine whether cannabinoids alter orexin neuronal activity as a primary or a secondary effect from changes in other sleep-related neuronal circuits.

In summary, we demonstrate that the CB1 agonist AM2389 produces a characteristic effect on sleep in male mice including suppression of REM sleep, increases in SWS, and consolidation of SWS. When the drug is given repeatedly over five days, some measures (REM, EEG spectral power) returned to near-baseline levels—suggesting tolerance—whereas others (SWS/Wake) showed rebound effects that were opposite to those seen after initial exposure. Following discontinuation of treatment, some of these effects (SWS/Wake) persisted, whereas for others (fragmentation of SWS, hyperactivity, EEG spectral power changes) a new pattern emerged that could indicate a spontaneous cannabinoid withdrawal in this species. The present study was limited to a single experimental protocol for AM2389 exposure; it is important to acknowledge that the impact of other regimens and cannabinoids needs further evaluation to determine if these effects are broadly applicable to this class of drugs. In addition, this study focused only on male mice; it is known that locomotor activity, body temperature, and sleep vary in female mice across the estrous cycle, which could potentially interact with many of the effects reported here (Missig et al., 2020). Nonetheless, these new findings suggest that cannabinoid withdrawal may produce characteristic patterns of sleep disruption that reflect their actions on the function of neuronal circuits that regulate sleep and wakefulness.